Embeddings

Use task_type embed to generate vector representations for spatial scenes and objects.

What it answers

What is the vector representation of this scene or object?

task_type: embed

What to send

input.cad_sceneobject

Scene to embed. point_cloud_scene also works.

input.object_idstring

Optional. Embed one object instead of the whole scene.

Every field on the request schema is still accepted; these are the ones this task depends on.

What comes back

embedding.vector

Float array for similarity search.

embedding.dimensions

Vector length, so you can size your index.

Example

Terminal

$ curl https://fourechelon.com/api/v1/responses \
    -H "Authorization: Bearer $FOUR_ECHELON_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
$     "model": "rowe-1.0",
$     "input": {
$       "task_type": "embed",
$       "cad_scene": { "schema": "four-cad.cad_scene.v1", "solids": [] }
$     }
$   }'

Worth knowing

Embeddings are for retrieval and clustering over your own corpus. Vectors are not stable across model versions — re-embed when the model id changes.